Tuning the Polarity of MoTe₂ FETs by Varying the Channel Thickness for Gas-Sensing Applications

Abstract

In this study, electrical characteristics of MoTe₂ field-effect transistors (FETs) are investigated as a function of channel thickness. The conductivity type in FETs, fabricated from exfoliated MoTe₂ crystals, switched from p-type to ambipolar to n-type conduction with increasing MoTe₂ channel thickness from 10.6 nm to 56.7 nm. This change in flake-thickness-dependent conducting behavior of MoTe₂ FETs can be attributed to modulation of the Schottky barrier height and related bandgap alignment. Change in polarity as a function of channel thickness variation is also used for ammonia (NH₃) sensing, which confirms the p- and n-type behavior of MoTe₂ devices.

1. Introduction

Since the discovery of graphene in 2004 [1], the two-dimensional (2D) layered materials have attracted significant attention for device applications owing to their unique physical properties and promising applications in nanoelectronic devices and circuits [1,2,3,4,5]. Due to lack of bandgap structure, graphene-based transistors are difficult to switch off, which is critical for electronic devices [6]. This limitation in graphene led researchers towards layered semiconductor transition metal dichalcogenides (TMDs), which are now seen as promising candidates for next-generation transistors due to their large variety of bandgap values [7,8], high charge-carrier mobility [9,10], and high on/off current ratio (~10⁶) [11]. Transition metal dichalcogenide (TMDC) compounds consists of one layer of transition-metal sheet (Mo and W) sandwiched between two sheets of chalcogenide elements (S, Se, and Te). Weak van der Waals interactions between the layers facilitate exfoliation of TMDCs crystals down to a single layer [3].

Among all TMDCs, MoTe₂ is the only material which can be grown in two phases: semiconducting (2H-phase) and metallic (1T’-phase). The theoretical bandgap for bulk and single-layer MoTe₂ in the semiconducting phase (2H) is 0.81 eV (indirect) and 1.13 eV (direct), respectively [12,13,14,15]. The bandgap of MoTe₂ is close to that of Si (1.1eV), making it an attractive candidate for optoelectronic devices with a range spreading from visible to near-infrared [1,11,16]. Recent literatures show that p- and n-type carrier injection can be obtained by work-function engineering of the contact metals [17]. The high work function of Pt and low work function of Ti metal contact can strengthen the p- and n-type conducting behavior [17]. It has been demonstrated that the field-effect transistor (FET) polarity can be electrostatically altered by dual top gate geometry [18]. The ambipolar behavior of MoTe₂ changing to n-type upon exposure to UV light for 2 h has also been reported [19]. It was also observed that unipolar p-type behavior of MoTe₂ flakes with Ti/Au metal contact and environmental oxygen can tune the device from ambipolar to unipolar p-type [11]. 2D layered nanomaterials are advantageous for gas-sensing applications due to their high surface-to-volume ratio, which facilitates surface reactions [20,21,22,23]. For chemiresistive-type FET gas sensors, electrical resistivity or conductivity are altered upon adsorption of target molecules on the surface of 2D nanomaterials [19,24,25]. It is a well-known fact that NH₃ acts as an electron donor (n-type doping), resulting in an increase of resistance for p-type semiconductors based on the charge-transfer mechanism [26].

In this study, we demonstrate that the change in polarity of MoTe₂ is a function of channel thickness. From our measurements, we observed that MoTe₂ FETs showed p-type, ambipolar, and n-type polarity with increasing flake thickness. MoTe₂ is intrinsically p-doped, but can also exhibit an ambipolar behavior [27]. To date, no clear evidence about unipolar n-type behavior of MoTe₂ FETs has been reported. From our experimental results of thin (~5.6–12 nm), medium (~12–45 nm), and thick (>45 nm) channel devices from tellurium tetrachloride (TeCl₄)-based transport-agent devices showed, p-, ambipolar-, and n-type conducting behavior, respectively. In this study we also highlighted the effect of channel thickness on gas-sensing application of MoTe₂ devices, which has not been mentioned in early literature [19]. The NH₃ gas-sensing results of MoTe₂ FETs further confirms the change in polarity as a function of flake thickness. No additional fabrication steps are required to obtain unipolar p- or n-type FETs. This electrical property of MoTe₂ can provide guidance to obtain p-type, ambipolar, and n-type devices, merely by tuning the channel thickness.

2. Materials and Methods

In this work, MoTe₂ flakes were exfoliated from bulk crystals grown by a chemical vapor transport (CVT) method. MoTe₂ (2H-phase) crystal was prepared by mixing powders of MoTe₂ and a small amount (ca 5 mg/cm³) of transport agent (TeCl₄), vacuum sealed in quartz ampoule for 140 h, and placed inside a furnace with temperature gradient. The temperature of MoTe₂ charge was maintained at 800 °C (hot zone), and the opposite end of ampoule was at 710 °C (cold zone). The ampoule was slowly cooled after seven days of growth [28]. The schematic diagram of the setup used for MoTe₂ crystal growth and an image of bulk crystal after CVT growth is shown in Figure 1a,b respectively. This method is known to produce pure 2H-phase of MoTe_2, as verified by X-ray powder diffraction (XRD), Transmission Electron Microscopy (TEM) [29], and Raman spectroscopy (Figure 1c).

Different thickness of flakes ranging from 5.6 nm to 60 nm were obtained from CVT-grown bulk MoTe₂ crystal, and transferred onto SiO₂/Si substrate by mechanical exfoliation (Supplementary Figure S1). The thickness of the SiO₂ layer was 285 nm. The blank SiO₂/Si substrate was patterned with design having numbers, symbols, and alphabets using photolithography. For fabrication steps involving photolithography, a bi-layer stack of positive photoresists LOR 3A and Microposit SPR 220.3 was used. LOR 3A was spin-coated at 4000 revolutions per minute (rpm), followed by a soft bake at 115 °C for 2 min. Next, Microposit SPR 220.3 was spin coated at 4000 rpm for 45 s and soft baked at 115 °C for 2 min. The samples were then exposed to UV illumination using mask-aligner (MA6 SUSS Microtec) and developed in Microposit CD 26A for 30 s, followed by a rinse in deionized water. Metal deposition in all fabrication steps was performed using an e-beam evaporator (Denton Infinity 22), and a lift-off process was performed by immersing the devices in remover 1165 at 80 °C for 30 min. These pattern marks were used for locating the desired MoTe₂ flakes after mechanical exfoliation. Prior to exfoliation, the patterned substrate was ultrasonically cleaned in acetone, iso-propanol (IPA), and deionized water (DIW), followed by oxygen plasma cleaning to remove the adsorbates from the surface [30]. Under the Olympus optical microscope, successful transfer of few layers of MoTe₂ flakes on SiO₂/Si substrate was mapped out to pre-fabricated marking points for a source/drain contact pattern. Fabrication steps shown in Supplementary Figure S2, was also used for patterning the source/drain metal contact of MoTe₂ devices. After fabrication, the devices were thermally annealed in a vacuum environment at 350 °C for 5 min to reduce the contact resistance between metal and MoTe₂ flake. Further, atomic force microscopy (AFM) was used to confirm the exact channel thickness and verify uniformity, absence of folds and cracks on the MoTe₂ flake under investigation.

After exfoliation, back-gate FET measurement was done using a Lakeshore probe station. The schematic of the back-gate FET connection is shown in Figure 2a. The silicon of substrate was used as a back-gate electrode and the SiO₂ layer acted as the gate dielectric in the back-gate FET measurement. These devices were characterized by output (Figure 2b) and transfer (Figure 2c) curves using back-gate FET measurement at room temperature. After the back-gate FET measurement, the devices were verified by NH₃ gas sensing for n- and p-type conducting behavior for the thick and thin channel. The gas sensing was performed by exposing 100 ppm of NH₃ concentration diluted with N₂ as carrying gas. N₂ gas is used to dilute NH₃ to the desired concentration at room temperature. We used N₂ as diluting gas due to its better recovery when compared with air environment. The gas sensing performance of the device was evaluated by applying 5 V between source and drain, and ground to the back-gate electrode.

3. Results

The flake-thickness scaling effect on MoTe₂ so far has gained less attention for the use in electronics compared with their thin counterparts. In this study, we have examined the conducting behavior of MoTe₂ FETs prepared using a TeCl₄ transport agent by CVT method. Using Raman spectroscopy with 532 nm laser source, we identified the lattice vibrational modes of mechanically exfoliated 2H-MoTe₂ flakes. The Raman spectra in Figure 1c exhibits characteristic A_1g at 170 cm⁻¹, E¹_2g at 235 cm⁻¹, and B_2g at ≈288 cm⁻¹ modes (the latter is active only in thin layers), which is in good agreement with studies reported in the literatures [31]. This verifies the 2H-phase of MoTe₂ and good crystalline quality used in FETs. Inset in Figure 1c shows the optical image of 4.6 nm fabricated MoTe₂ FET. Optical imaging technique has been popular in the rough estimate of flake thickness [32].

To understand the channel-thickness effect of FET transport characteristics, we measured the output (drain current (I_ds) vs. drain voltage, (V_ds)) and transfer (I_ds vs. back-gate voltage (V_bg)) characteristics using Lakeshore probe station. The schematic diagram of circuit connection for back-gate FET measurement is shown in Figure 2a. In Figure 2b, the output characteristics of the 10.6 nm MoTe₂ FET shows an increase in conduction as gate voltage decreases from +40 V to –40 V, showing that the majority of carriers are holes, since the channel is entering the ON state with a negative gate voltage. Contrarily, the output characteristic of the 56.7 nm MoTe₂ device, shown in Figure 2c, shows that the majority of carriers in the channel are electrons (ON state with positive gate voltage) as gate voltage decreases from +40 V to –40 V. Clearly, p-type and n-type transport behavior is observed for 10.6 nm vs. 56.7 nm MoTe₂ FETs, respectively. The nonlinear behavior of the output curve in Figure 2b,c can be attributed to a small Schottky barrier at the metal/semiconductor junction due to difference between the work function of Ti (4.33 eV) and the electron affinity of MoTe₂ (4.3 ± 0.1 eV) [17,33].

We further studied the transfer behavior of the FETs at V_ds = 2 V. The transfer characteristic results showed that thick, medium, and thin channel are n-type, ambipolar, and p-type, respectively. The gate leakage current (I_gs) measured in the pA range (negligible). The maximum on/off current ratio obtained was ~1 × 10⁴, which increases with decreasing channel thickness. Figure 3 shows transfer curves for varying channel thickness for different metal contacts. Multiple devices with various metal contact (Ti/Au, Cr/Au, and Pd/Au), were fabricated which reproduced similar transfer characteristic as a function of MoTe₂ channel thickness [34]. The metal work functions (ϕ_m) of Ti, Cr, and Pd are 4.3 eV, 4.8 eV, and 5.1 eV, respectively. At least 10 devices from each metal type were fabricated and tested to reproduce the transfer curves to verify the change in polarity based on channel thickness. From transfer curves, we observed that MoTe₂ FETs showed p-type behavior for very thin channels, ≈5–15 nm and ambipolar behavior for medium channel thicknesses, from ≈15 to 50 nm. For channels thicker than ≈50 nm, FETs showed n-type unipolar behavior. Previous studies on TMDC (such as WSe₂, MoS_2, and MoTe₂) have presented the effect of channel thickness on various transport properties in FETs [35,36,37,38] , but a wide range of channel thickness (~5–60 nm) and its effect on polarity of MoTe₂ devices has not been highlighted until now.

In reference to n-type conducting behavior, the n-type doping has been observed in TMDCs when TeCl₄ is used as the transport agent in CVT growth [39] and during post-growth chloride molecular doping of TMDC compounds [40]. In this work, we assume that the Cl doping is playing a dominant role for n-type behavior in thick channel devices. In thin-channel FETs, conductivity switches to p-type, indicating the diminishing of Cl doping. We attribute it to the increase in surface defects and adsorbates in ultra-thin layers, which is also exemplified in atomically thin MoS₂ FETs [35]. In addition to the possible effect of reduced MoTe₂ channel thickness on n-type doping efficiency, we also speculate that the polarity switching from n- to p-type in thinner layers may be caused by the modulation of Schottky barrier height (SBH) and corresponding band alignment and band-bending at the metal/MoTe₂ interface. The effective barrier height for Ti, Cr, and Pd contacts are 41.1, 40.3, and 10.2 meV, respectively [41,42].

We also investigated the field-effect carrier mobility (µ_FE) of fabricated devices extracted from the transfer curve of Figure 3a and MoTe₂ devices with Ti/Au as metal contact. Mobility, µ_FE of the back-gated MoTe₂ FETs is defined as follows:

µ_FE = (dI_ds/dV_bg) {L/(W C_OX V_ds)}

(1)

where C_OX (=ε₀·ε_r/d) is silicon oxide capacitance per unit area, d is the oxide thickness, ε₀ and ε_r are the relative permittivity of free space and the relative permittivity of SiO_2, respectively, L and W are the channel length and width, respectively, V_ds is the drain-source voltage, and (dI_ds/dV_bg) is the inverse slope of transfer characteristic in the linear region of the ON state. The mobility increases with increasing channel thickness, as shown in Figure 4. A thicker channel has higher mobility compared with thinner MoTe₂ devices. The influence of channel thickness in MoTe₂ devices can be associated to Coulomb scattering and quantum confinement, as reported in early literatures [43,44]. Coulomb interactions weaken the scattering of carriers, resulting in higher mobility for thicker FETs compared with thinner counterparts, which was also demonstrated for both MoTe₂ [35] and MoS₂ [45] FETs.

To further confirm the change in polarity of FET devices, their response in the presence of gas molecules was measured. Figure 5 shows the gas-sensing setup used for NH₃ sensing in this work. Nitrogen (N₂) as diluting gas is connected to MFC2 and mixed with NH₃ target gas (connected to MFC1). Both MFC1 and MFC2 are mixed in MFC mixer for the required concentration. The output of MFC1 and MFC2 is connected to MFC mixer for the required concentration of NH₃ to flow in the gas-sensing stainless steel chamber for measurement. The MFC mixer is also connected to a pressure controller used to vent out extra gases. The MoTe₂ device is packaged on ceramic chip carrier (inlet of Figure 5) and placed inside the stainless steel chamber, and a source/drain is connected to probes of a National Instrument (NI) for data collection. The gas sensing data from the NI instrument is analyzed using LabView program by transient curve.

The schematic of NH₃ sensing on n- and p-type is shown in Figure 6a–c. Figure 6a shows the schematic of process flow for NH₃ gas sensing. We selected NH₃ as the target gas since it is a strong electron donor and will produce opposite behavior in n-type (Figure 6b) and p-type (Figure 6c) devices. When NH₃ is adsorbed on the n-type MoTe₂ (Figure 6b) surface, an increase in current is observed due increase in electron charge carriers on the MoTe₂ surface. When NH₃ gas is switched off, NH₃ molecules are desorbed from the surface and the current value of pristine-state MoTe₂ surface is obtained. Similarly, when NH₃ is adsorbed on the p-type MoTe₂ surface, a decrease in current value is seen, due to a decrease in hole-charge carriers on the MoTe₂ surface, as shown in Figure 6b. A single pulse of 100 ppm concentration of NH₃ was introduced to the MoTe₂ devices, and the current response of devices was measured.

Figure 7a,b shows the gas-sensing response of 100 ppm NH₃ of 60 nm and 6.1 nm MoTe₂ flakes. We were able to detect the NH₃ concentration as low as 1 ppm for thin MoTe₂ channels. We couldn’t go below 1 ppm due to a limitation of the mass-flow controller. The lowest concentration detected by thicker channel devices was 100 ppm. Hence, for comparison purpose of n- and p-type MoTe₂ devices, 100 ppm NH₃ sensing was used in this work. The thick (60 nm), n-type MoTe₂ FET showed an expected increase in current, and the thin, p-type MoTe₂ FET showed a corresponding decrease in current. The response is calculated using Equation (2) for Figure 7.

Response (%) = [(I_gas − I_baseline)/I_baseline] × 100

(2)

where, I_gas and I_baseline are currents when NH₃ is on and the baseline current, as shown in Figure 6b,c. This experiment verifies the change in polarity in MoTe₂ flakes due to variation in channel thickness. Thin flakes show higher sensitivity (~12 %) compared with thick flakes (~5%) due to high surface-to-volume ratio. The current values of NH₃ sensing by thin and thick MoTe₂ flakes is shown in Supplementary Figure S3. The low value of sensitivity and baseline drift in devices can be due to the limitations of the custom-built gas-sensing setup. In this work, we showed that electrical properties of MoTe₂ are dependent on flake thickness, which is also verified by NH₃ sensing of thin and thick flakes. No extra fabrication steps are required to obtain the n- or p-type conduction behavior in MoTe₂ FETs. We predict that with further experiments and modifications in our gas-sensing setup, a similar change in polarity of MoTe₂ device as function of channel thickness can also be shown for NH₃ sensing in an air environment.

4. Conclusions

In this work, we demonstrated a simple and effective way to fabricate p-type, ambipolar, and n-type MoTe₂ FETs just by tuning the channel thickness. FETs were fabricated from mechanically exfoliated flake from CVT-grown single-crystal 2H-MoTe₂. The electrical properties of devices prepared from 2H-MoTe₂ flakes were examined for varying thickness (~5–60 nm) using the output and transfer curves. FETs showed polarity switching from n-type through ambipolar to p-type with decreasing channel thickness from 60 nm to ~5.6 nm. The n-type transfer behavior in thick-channel MoTe₂ FET is attributed to chlorine doping from the TeCl₄ transport agent used in CVT growth of bulk crystals. The switching of polarity by thinning the FET channel may be associated with the increasing role of surface states in ultra-thin layers, which can influence charge-carrier concentration by modulating the Schottky barrier height between metal and semiconductor interface. This study also showed the NH₃ gas-sensing application of p- and n-type MoTe₂ devices, indicating the change in polarity due to channel thickness.